COMPILATION AND APPLICATION OF IDE-JETRO S INTERNATIONAL INPUT-OUTPUT TABLES

Transcription

1 OMPIATION AND APPIATION O IDE-ETRO S INTERNATIONA INPUT-OUTPUT TAES o Meng International Input-Output Analysis Group Institute of Developing Economies, ETRO, apan bo_meng@ide.go.jp Yaxiong Zhang Department of Economic orecasting, State Information enter, hina zhangyx@mx.cei.gov.cn Satoshi Inomata Director, International Input-Output Analysis Group Institute of Developing Economies, ETRO, apan satoshi_inomata@ide.go.jp ASTRAT International input-output tables have been considered one of the most useful data sources in economic analyses. Since these tables provide detailed information of international production networks, they have attracted significant attention recently in the research areas of spatial economics, global value chains and trade in value added related issues. IDE-ETRO has more than 40 years of history in the construction and analysis of international input-output tables. This paper gives an easy-to-understand introduction on IDE-ETRO s international projects including the construction of Asian International Input-Output Tables and Transnational Interregional Input-Output Table between hina and apan. In order to help users understand the characteristic features of our tables, this paper also shows some application examples by using these tables. Keywords: International input-output, Production networks, Economic interdependency

2 . INTRODUTION Input-output (IO tables show the flow of goods and services between all the individual sectors of a national economy over a given period of time (normally, one year. They also describe the sale and purchase relationships between producers and consumers within an economy. As such IO tables form the core of the system of national accounts and also have become an important instrument to guarantee and enhance the quality of other statistics. esides playing the role in descriptive purpose in statistical areas, IO tables have also been widely used for analytical purposes in applied economic research and policy analysis. or example, IO tables can be used for the analysis of production, cost structures and productivity; analysis of prices; analysis of the structure of capital formation, final consumption, exports, etc.; analysis of imports of energy required; impact analysis of new technologies; sensitivity analysis of the effects of changes in tax. If time series of IO tables are available, it is possible to perform structural decomposition analysis of economic growth or to build econometric IO model. urthermore, IO tables have been a prerequisite for building GE (omputable General Equilibrium models. Due to the importance of IO tables, today there are more than 0 countries around the world that have compiled their own national IO tables and published them officially. However, just using national IO table, it s difficult to illustrate how a target country links or interacts with other countries by the way of international production networks. This is mainly because that national IO tables cannot provide the information about how a country s exports are used by another country as intermediate inputs or how the imported goods are produced in its partner countries. In response to the recent development of spatial economics (New Economic Geography, New new trade theory and global value chains related issues, it is increasingly recognized that the international IO table can play an important role in related analysis areas. The construction of international IO tables is not a simple process of just linking every country s national IO table with international trade statistics together. It s exactly a very artistic practice. When compiling an international IO table, compliers have to face a lot of difficulties and challenges. This is mainly due to the fact that the original national tables are basically compiled by individual national statistical offices with relatively independent or different definitions, sector classifications and compilation methods. Such fact also reflects the difference of original statistical system, economic idiosyncrasies and availability of data across countries. Therefore, the first hard work for compilers is to harmonize every country s national IO table to fulfill the common standards of the compilation of international IO table. In addition, when linking the national IO tables with international trade statistics, there is no guarantee that these two data sources can be consistent well originally. Therefore, a number of adjustment processes have to be done carefully based on some assumptions and supplementary data sources. urthermore, the quality of compiled international IO tables depends on not only the original data collected, but also the compilation and estimation methodologies used. inally, compiling international IO tables is not only a time-consuming and money-consuming practice, it also needs a number of experts from both statistical and economic sides to work together and exchange considerable amounts of information, 2

3 ideas and technical expertise. Therefore, better management of international communication of different member countries is also a very important factor during the construction of international IO tables. The Institute of Developing Economies (IDE ETRO has a long history in the field of construction of international IO tables. Its main products include a series of Asian International Input-Output (AIIO Tables for the years of 95, 95, 995, 2000, 2005, 200 2, and the recent newly developed Transnational Interregional Input-Output (TIIO Table between hina and apan for the year of 2000 under the collaboration of hina s State Information enter (SI. The main purpose of this paper is to provide an overall view of IDE s international IO projects. In order to help users understand the characteristic features of IDE s projects well, this paper also shows some analytical application examples by using these tables. The paper is structured as follows: in section 2 and 3, the details of AIIO and TIIO projects are introduced respectively. Every section includes the project history, background, organizations involved, data source and methodology. In section 4, some application examples are introduced. or the AIIO tables, we mainly show some analysis results of vertical specialization measured by the AIIO tables and transnational employment trade by using the TIIO table. 2. OMPIATION O THE ASIAN INTERNATIONA INPUT-OUTPUT TAES As a spatial extension of national IO techniques, the international IO table has become a powerful tool for economic analyses. IDE for the last 40 years has been making a lot of efforts to construct various international IO tables in collaboration with statistical offices and research institutions of Asian countries. Now facing the rapid growth of the hinese economy and the deepening economic interdependency in the Asia-Pacific region, IDE s AIIO Table has been an indispensable apparatus for the analysis of Asian economy from a spatial perspective. This section provides some fundamental information about IDE s AIIO international project. 2. History of IDE s Asian International Input-Output projects The whole history of IDE s AIIO project can be divided into four phases. In the first phase (before 93, some pioneering works had been done. In 964, IDE s Watanabe (966 proposed the idea of using international IO models to analyze the North-South trade problem. ollowing this idea, in 965 IDE developed an international IO model covering six regions: North America, Europe, Oceania, atin America, Asia, and apan. A preliminary version of the 2005 AIIO table has been completed in 20, but not officially published. The official AIIO data for 2005 will be released on March In order to response to a special requirement on studying the economic impact of 200 financial risis on Asia-Pacific region, a very preliminary version of the 200 AIIO table has been complied, but not officially published. 3

4 In 966 and 9, IDE constructed international IO models for ten Asian countries. Yet the IO tables compiled in these studies were subject to a number of limitations; i.e., estimation techniques were too simple, the industrial sector classification was too crude, and so on. In the second phase (93-9, IDE launched development of a comprehensive international IO table to explore the situation of inter-industrial transactions among East Asian countries. In 93, the Republic of Korea and the five ASEAN countries plus the United States were chosen to be the endogenous member countries, as these countries have close economic relationships with apan. Had all the countries compiled their national IO tables at the same referential year, the project would not have been so laborious. However, Indonesia, Thailand, and Singapore had not constructed any IO tables by that time. Also, IDE was not sufficiently experienced in compiling a comprehensive IO table. Thus, the project had to begin with two preliminaries: one was to construct national IO tables for these three countries, and the other was to compile bilateral IO tables with the countries for which the national tables were already available. Under this project, three national IO tables (Indonesia for 9, Singapore for 93, and Thailand for 95 and three bilateral IO tables (Korea-apan for 90, U.S.A.-apan for 90, and Philippines-apan for 90 were constructed in collaboration with the national statistical offices and research institutes of the countries concerned. rom 9, IDE entered its third phase (9-92 of the IO project, with the aim of constructing the 95 multilateral IO table among the ASEAN countries, apan, Korea, and the U.S.A. This project went along with the following three steps: ( estimation of national IO tables for the countries that did not have 95 national tables, (2 construction of 95 bilateral IO tables for the countries that had already compiled national tables by the time, and (3 construction of the 95 multilateral IO table. irst, existing tables had to be updated to the year 95 for Malaysia, the Philippines, Singapore, and the U.S.A. Next, the 95 bilateral IO tables for Indonesia-apan, for Thailand-apan, and for Korea-apan were constructed. inally, these national and bilateral tables were linked together under a single multilateral IO table for 95, which was completed in 93. The 95 multilateral table has been used for various analyses of East Asian industrial structure, and the table became the prototype for the subsequent international IO projects. After the completion of the 95 international IO tables, IDE soon launched the construction of the 95 international IO table, to cover more Asia-Pacific countries. Since hina commenced an Open-Door policy as one of its key development strategies it has rapidly increased its external trade with the United States, apan, and others. hina plays an important role in the Asia-Pacific region, not only in providing a gigantic market but also in receiving investment from the neighboring countries. Thus, hina and Taiwan were covered in the 95 multilateral table, making it even more comprehensive than the previous 95 version. Since then, IDE has successfully completed the multilateral tables every five years, providing powerful analytical tools for dynamic structural changes in the Asia-Pacific region. 4

5 2.2 Organizations involved in the AIIO projects IDE s international IO projects involve many participants (organizations from different economies. Most of them are the governmental or semi-governmental statistical offices and institutions. or example, more than 0 experts from ten different economies have contributed their efforts directly and indirectly to the compilation of the 2000 AIIO table. The organizations involved in the 2000 AIIO projects are shown below. hina: State Information enter Indonesia: Statistics Indonesia Korea: The ank of Korea Malaysia: Department of statistics The Philippines: National Statistics Office Singapore: usiness Research onsultants Taiwan: Taiwan Research Institute Thailand: The National Economic and Social Development oard apan: Applied Research Institute, Inc. apan: IDE-ETRO In addition, some statistical experts from Gifu Shotoku Gakuen University and the Ministry of Economy, Trade and Industry of apan also provided their help on this project. 2.3 ompilation procedure and methodology of the AIIO tables The AIIO table includes ten endogenous economies (see Annex and 6 sectors. Roughly speaking, the compilation process of the AIIO tables goes through three distinctive phases. What follows is a step-by-step illustration of how the AIIO Table is compiled. The first part presents a description of the format adjustment for every constituent national table based on the general survey on national tables, which was conducted by IDE in 2003 and 2004 in order to establish a common rule for the format adjustment of the tables. The second part briefly explains construction of the system of sector concordance, followed by a brief introduction of estimation methods for supplementary data. inally, the linking procedure is illustrated, with detailed explanations on the manual balancing and reconciliation work Adjustment of presentation format Despite the fact that IO tables constitute the central apparatus of the System of National Accounts, each national table of an individual country exhibits more or less different features and characteristics, reflecting the country s economic idiosyncrasies and availability of data. Such a variety in the form, however, poses a practical difficulty when compiling international IO tables. or even though the international table is composed of the segments taken from each national IO table, the interpretation of the data should be mutually consistent and comparable for any part of the whole. Accordingly, one of the most complicated, nerve-racking tasks of compilation is the adjustment of national tables to conform to a common format. In general, it is the 5

6 detailed, information-rich table that has to compromise with less-detailed ones, as the other way round would require a costly (yet often unrewarding effort of obtaining supporting data. Therefore, there always exists a trade-off between the level of uniformity and the level of information, and hence careful and thorough consideration is called for in making adjustment rules. This section reports on the general survey on the characteristic features of national tables of AIIO member economies. The survey was conducted in the period of , in order to construct the basic information reserves for designing the AIIO common format and adjustment rules. To our knowledge, such an extensive and detailed survey on national tables has never been carried out, and we believe that no institution but the IDE, with a history of significant cooperative relationships with IO experts of various Asian economies, would be able to make such a substantial survey possible and successful. In our survey, a questionnaire was carefully designed so as to capture every important aspect of an IO table. The questions are grouped under seven broad categories, namely: enchmark-year and recording principles, Availability of national tables and supporting tables, Valuation, orm and coverage, Special treatment, Public or semi-public sectors, Response to the 993 SNA. ased on the survey results, the major findings can be summarized as follows: ( Similarity to the apanese I-O table In igure, the degree of similarity to the apanese IO table is illustrated. The horizontal axis is the level (number of industrial classification, while vertical axis concerns presentation format, giving the percentage rates of the number of questions in the questionnaire to which the country gave the same answers as apan's. (The rates are calculated against the sum of valid answers only. The diagram shows that the most similar table of all is the Korean IO table, as its industrial classification has just one sector difference with that of the apanese table, and the rate of the same answer is more than 0%. Then, we can identify the second group, including Indonesia, the Philippines, Thailand and Malaysia. Not to mention about the Korean table, there is no wonder for these tables (except Malaysia to show high degree of similarity to the apanese table, since their national IO projects are known to have been initiated and conducted under the advice and support of apanese IO experts. The US table is indicated as having some degree of similarity, but in the survey result it is observed that many answers remain to be unknown, so that no conclusive evaluation can be made against this table (although it is true that the classification difference is the second smallest after the Korean table. 6

7 Presentaion format igure : Similarity to the apanese I-O table korea 0.65 indonesia philippines malaysia thailand usa taiwan 0.4 china singapore Industrial classification apanese I-O Table olumn=405 Table : Responsiveness to the 993 SNA Rank ountry rate * PHIIPPINES 0.54 USA THAIAND KOREA APAN SINGAPORE INDONESIA MAAYSIA HINA TAIWAN * The percentage rates of the number of questions in section of the questionnaire to which the country gave the answer that follows the SNA recommendation.

8 The third group, which is least similar to the apanese table, includes Taiwan, Singapore and hina. In addition to the dissimilarity of the format and of the level of industrial classification, the benchmark years of these national tables differ from that of apan, i.e., with 0 or 5 in the last digit of the year. So the official tables had to be updated to the year 2000 with the help of some estimation methods like RAS algorithm, and this will further disturb the accuracy of the tables. The same is true for the national table of the United States. (2 The responsiveness to the 993 SNA The System of National Account is a comprehensive guideline for compiling national statistical data. If properly followed, the resulting statistics will be mutually consistent and internationally comparable. The latest version of the SNA, the 993 SNA, underwent an extensive revision of its predecessor, the 96 SNA, to bring the statistical notions and methods up to date. IO tables (or more precisely supply and use tables, which constitute a core apparatus of the System, didn t remain unaffected, and many countries including our project partners have put every effort to make their tables accordant to the new scheme. The survey result (Table 2 shows that the most responsive countries are the Philippines and the United States, yet again one must be careful about the result on the US table as it contains a number of unknowns. The Thai IO table comes next, followed by the Korean and apanese table. Although the Korean table and apanese table ranked the same, the former can be evaluated higher as it already succeeded in introducing one of the most challenging schemes in the 993 SNA, i.e., the inancial Intermediary Services Indirectly Measured (ISIM. On the other hand, it is rather surprising to observe that Singapore and Malaysia came to low ranks, as these national tables are known to have followed the previous 96 SNA schemes quite extensively. (3 The areas of conflict inally, we shall briefly look at the areas of conflict where each country s treatment is not in line. The most prominent example is the treatment of Scraps and y-products. There are normally four adjustment methods for this problem. Each of them has both advantages and disadvantages, and the member countries employed the various schemes in quite an uncoordinated fashion. In the absence of supplementary information on the generation and use of scraps / by-products, it is not possible to convert from one scheme to another, making it difficult to reach a common agreement on the adjustment method. The second area of conflict is about the treatment of imputed interest. The previous 96 SNA recommended that the output of imputed interests (= the difference between the interests receivable and the interests payable should all go to intermediate transaction, not to final demand. The countries like apan, Singapore and Malaysia strictly follow this stipulation, while other countries tables have output in final demand as well. The introduction of ISIM under the 993 SNA may provide an integrated guideline for this issue, but so far no member country except Korea is successful in introducing this new scheme.

9 HINA INDONESIA APAN KOREA MAAYSIA TAIWAN PHIIPPINES SINGAPORE THAIAND U.S.A. The last prominent area of conflict is the treatment of inventory. The related question in the questionnaire is: Suppose that a car industry (demand-side sector purchased a set of tires (supply-side sector but did not use them this time. How does this input enter in the table? Most of the countries answered that the input should be recorded at the intersection between Tire (supply-side industry and hange in Stocks, but some countries like hina, Taiwan and Singapore answered the opposite, i.e., at the intersection with ar (demand-side industry. Singapore gave an explanatory comment on this. It treated this input as a stock of car since tires are regarded as a <work-in-progress> of a car. It is quite surprising to find out that even the very basic economic concept like an inventory is in fact yielding to different interpretation among countries. Table 2: Different features and characteristics of the AIO member economies. onversion of valuation. asic price to producer's price X.2 Private onsumption Expenditure X X X.3 Export vectors X X.4 Import matrix/vector X X X X X 2. Negative entries X 3. Dummy sectors X X X X X X 4. Machine-repair X X X X 5. inancial intermediaries X X X X 6. Special treatment of import/export 6. Water transportation X 6.2 "Pure import" of gold X 6.3 Re-export X 6.4 Telecommunication X. omputer software products X. Producers of government services X X Using the results of general survey on the characteristic features of national tables of AIIO member economies, the main adjustment targets for each national table can be determined (see Table Preparation of sector-concordance and supplementary data efore linking every part provided by different member countries, we need to ( harmonize each national IO table under the common sector classification of the AIIO table; (2 prepare trade related information; (3 utilize the information getting from a 9

10 special survey on imported commodities. This section gives a brief introduction on each preparation procedure. ( Preparation of sector concordance Each national table has its own industrial classification. In the case of the benchmark tables for the 2000 AIIO table, the number of industrial sectors ranges from 9 for the Malaysian table to 5 (row for the apanese table. The weight of the industrial category also differs. The countries with large agro-based economies have relatively detailed classification of agricultural sectors, while industrialized economies give more comprehensive coverage to manufacturing sectors. As such, the sector classification reflects the characteristics of the economy concerned, and a precise conversion system that bridges national codes and AIIO codes is absolutely essential for the compilation of consistent international IO tables. The system of sector concordance has a treelike image, where AIIO classification (the broadest category rests on the top, and each AIIO code corresponds to one or several national codes. The national codes are subcategorized into the Harmonized System of oreign Trade Statistics, which may be further converted to SIT, another classification system for the trade data. If the concordance system has such a clear-cut tree structure, the aggregation of national tables into AIO classification poses no difficulty. The problem arises when a national code is associated with more than two AIIO codes. or example, Singapore s national code SIO092 and transport equipments corresponds to both AIO055 Motor vehicles and AIO056 Motorcycles. Here, the sector splitting of the national IO table is called for before the aggregation procedure. (2 Supplementary data or the compilation of international tables, the following supplementary data should be prepared by each country at AIIO sector classification. (a Import data by commodity and by 2 economies of origin 3 ; (b Export data by commodity and by 2 economies of destination; (c Import duties and import commodity taxes by commodity; (d Domestic trade margins and domestic freight transport costs (TTM on exported goods by commodity; (e International freight and insurance by commodity and 0 endogenous economies of origin; (f Other relevant information. The import and export data can be directly constructed from the oreign Trade Statistics, with the help of the HS (or SIT - national IO - AIIO sector concordance. The data on import duties and import commodity taxes, on the other hand, are independently presented in the original national IO tables in most cases, but if not (as in the case of the U.S. table, they must be also collected from the oreign Trade Statistics. 3 economies: project member economies plus Hong Kong, EU, the Rest of the World 0

11 The data of TTM on export comes from the supporting tables of the national IO tables. Ideally, those levied on exported goods (for the delivery from factories to ports should be used, but if they are not available from the table the average figures of the TTM matrices can be used as proxies. inally, the data on international freight and insurance are collected from the oreign Trade Statistics, where available. Yet, because not all countries have these data, we use gravity model to estimate the missing information. In our model, two measures of distance are calculated, i.e., the shipping-route distance and the straight-line distance. The shipping-route distance is taken from the Distance Tables for World Shipping (SE 93 in which the distances between major ports are reported. The straight-line distance, which can be regarded as an analogue of the air-flight distance, is calculated between commercial centers of the countries concerned. Of these two measures, the one that better explains variation in international freight and insurance rates is employed. (3 Getting information from a special survey on the use of imported commodities. It should be noted, the quality of import matrices plays a critical role in determining the accuracy of the international IO table. In order to increase the accuracy of import matrices, a special survey on imported commodities has been done in the current AIIO project. The main purposes of the survey include: (a to identify using industries of the imported commodities by origin country; (b to determine the value/rate of the international freight and insurance on each imported commodity; (c to determine the value/rate of import duties and commodity taxes levied on each import commodity. The respondent of the survey will be the establishments that import the commodities (manufacturers, trading firms, etc. as they are considered to possess the information on amount imported by country of origin and their distribution amount to domestic industries. The survey is basically carried out as an independent sample survey. Also it may be conducted as a rider survey attached to other official surveys, which is more efficient and comprehensive. (The sample form of the questionnaire in order to collect the information described above is presented in Annex 2. arrying out the special survey described above accompanies several problems. irst is the feasibility of the survey. It is difficult for some countries to conduct the survey, owing to the lack of resources (funds, personnel, connections with related authorities and firms, knowledge, etc.. or countries that the survey is infeasible, it will be required to look for some alternative solutions. One possible alternative is to modify the import matrices by referring to other countries survey results. Second is the sampling issue. Even if the survey can be carried out, it is not easy to collect the reliable information. or instance, the samples should be selected in order to represent the characteristics of the industry appropriately. However, identifying the typical samples that appropriately reflect the distribution structures is not easy.

12 Third, it may also be difficult on whether or not the distribution structure can really be determined, even if samples are chosen appropriately. This problem has two different aspects. One aspect is the difficulty to determine the final users of imported commodities by country of origin. As discussed above, the imported commodities are usually delivered to the final users through the wholesale and retail agents. The respondent to the questionnaire, the importing firm, may not have the information on the final users if they sold their imports to the domestic wholesalers or retailers. The other aspect of the problem is that it may be difficult to determine the amount sold of each imported commodities even though the final users can be identified. This may occur if the survey year is distant from the reference year that the respondent cannot trace the transaction records as they may not keep the detailed information inking and balancing An international IO table is not just a patchwork of the pieces taken from national tables, but it is a product of careful utilization of supplementary data and manual reconciliation or fine-tuning work. This section gives a brief description on the linking and balancing work of the AIIO tables. ( inking of national IO tables All the pieces of each member country prepared in the previous steps are linked together to the one big table as shown in Annex. The basic idea of linking is to replace the export vector by the import matrix of the trading partner. In other words, we first split the import matrix of each country s national IO table to all its partner country and then link these tables row-wise. At this stage, the valuation of imports in each country s national IO table is also converted from the.i.. price to the producer s price by using the data of international freight and insurance, and domestic transport costs and trade margins compiled in the previous steps. (2 alancing and Reconciliation of the whole table The final step of compilation is the manual balancing and reconciliation work, following the linking of all the pieces provided so far. The table is balanced with respect to the input composition, but total demand is not necessarily consistent with total supply for each country at this stage. Such an imbalance (discrepancy stems from various reasons. The first one is the inconsistency between each country's sector classifications. Though each country is required to make its own code concordance from HS code to AIIO sector classification, the possibilities of differences in statistical concept still exists. The second one is because of entrepot trade. This type of trade is always counted in different ways by trade partners. or example, in the case of hina, export via Hong Kong to the U.S.A. may be counted by the U.S.A as import from hina. In the case of Singapore, where international trade is extremely large compared to the scale of its economy, and there is a large volume of entrepot trade, there are especially large statistical discrepancies in its international trade matrices. 2

13 Number igure 2: Procedure of balancing and reconciliation Start (a Total check by published data sources (b ompute the T s error rates (c T s error rate > specified criterion Yes (d alculate the error in Trade Statistics No End (e Identify the reasons for the errors (f Make adjustment policy/instruction and adjustment card (g Run adjustment program igure 3: Distribution pattern of T error T' error rowwise > 30% T' error rowwise > 0% T' error rowwise > 5% T' error rowwise > % hina Indonesia apan Korea Malaysia Taiwan Philippines Singapore Thailand USA ountry/region 0 3

14 In order to rationally and efficiently decrease the discrepancies generated through the linking process, the procedure shown below (see igure 2 is employed in final reconciliation of the AIIO table. (a Initially, we use the linking results to summarize the transactions among the industries of all countries and compile an AIIO table that there is only one sector per country. Then it becomes easy to check whether or not the present data in the AIIO table at the national level are consistent with the published data sources, such as the GDP statistics published by each country or the IM statistics. Through the above checking, we can have the overall information on the preliminary linking results. (b In order to determine the size of the final adjustment in detail, we calculate the error rates of T row-wise by sector for each country. igure 3 shows the distribution pattern of the absolute rate of T error at different levels for the 2000 AIIO table. The vertical axis represents the number of sectors in which T's errors are larger than the specified levels. Obviously, hina, apan, and the U.S.A. have relatively smaller numbers that are counted at each level. On the other hand, Indonesia, Malaysia, the Philippines, Singapore, and Thailand have relatively larger numbers. Korea and Taiwan show the similar pattern. The distribution pattern shown in igure 3 not only depends on the economic scale but also relates to the statistic system of each country. onsidering the large scale of the AIIO table and the distribution pattern of error rates, any sector with relatively large rate of T error (5% is determined as the target for our adjustment. (c Though 5% is determined as the criterion for the preliminary adjustment, considering that positive errors and negative errors may offset each other row-wise, we have to investigate the structure of the errors in detail. As stated in the previous section, the AIIO table is mainly based on the import matrices of each country s national IO table. However, there is no guarantee that a country s import statistic can be completely consistent to its partner s export statistics. In order to discuss the structure of the error in detail, we calculate the difference between a target country s imports and its partner country s export for every commodity. When looking at this information, the structure of a target country's T error row-wise becomes easy to understand. This also offers us information about which sectors and which countries should be the main targets for adjustment. (d According to the analysis of bilateral trade information obtained from preliminary linking results and careful investigation of the HS-AIO code concordance, the majority of errors can be identified. Then the adjustment policy will be determined. ased on detailed discussion on the adjustment policy among related member staffs who are in charge of different target country, an adjustment card can be produced. (e The adjustment cards are used as input file in our adjustment program. asically, the adjustment is merely executed on the import matrices, and it vertically moves the same figure from one sector to other sectors. This means that the T balance can be maintained column-wise. 4

15 The above procedures (a - (e will be repeated until the results satisfy the specified criteria. Additionally, spot-check is conducted at the end of the adjustment. This is to spot out any unnatural entries in the table that might have been brought in during the process of the adjustment. or example, the output of electricity, gas & water supply or some other service sectors is not supposed to have any cells along ixed apital ormation or hange in Stocks. Any of such mis-tabulation should be adjusted properly. 2.4 hallenge towards the future project of international IO tables Given the increasing economic interdependence across countries caused by the extension of globalization and regional integration, international IO tables have been considered a very useful data source for the analyses of production networks, international fragmentation production, global value chains and so on. In response to the increasing attentions and requirements from many policy makers and researchers, there are a lot of challenges reminded towards the future project. The first challenge is about the time lag of publication. IDE compiles the AIIO tables every five years. However, there is always more than five-year time lag between the benchmark year and reference year. Since most countries construct their national IO table per five years, and also the benchmark year across countries are different, this makes it difficult to speed up the process of linking every countries parts together in time. If the statistic system in much more Asian countries can switch to or follow the SUT (Supply and Use Table, annual national IO table can be easily estimated. This will help the compilation of international IO table become speedy. The second challenge is about how to minimize the discrepancy arising from the linking process. As mentioned in the previous sections, the most important reasons causing the discrepancy are ( the inconsistency of export/import figures between national IO table and international trade statistics; (2 the mirror problem in bilateral trade statistics caused by the treatment re-export and re-import; (3 the different treatment of valuation between export statistics (.O. and import statistics (.I.. One of the possible solutions on the above problems is to apply the recent UN s E (road Economic ategories classification to the current trade statistics 4. Under this classification, trade data can be grouped into different end-use categories, like intermediate goods, final consumption goods, capital goods and so on. This can improve the precision of allocation of bilateral trade data when linking the national IO tables. In addition, according to the new recommendations for IMTS (International Merchandise Trade Statistics proposed by UNSD, imports on the.o. basis in addition to the standard.i. valuation is expected to be published in the near future. This may help us make much more consistent bilateral trade data. inally, the re-export statistics by country of origin and destination should be an important source in the solution of the mirror problem happened in trade statistics. 4 or the recent information on end-use and industry based trade data, one can refer to the OED s work (see Zhu,Yamano and imper 20 5

16 The third challenge is about the valuation. The AIIO tables are at producer s prices. It is no doubt that the most preferable valuation concerning the requirement of economic analysis is basic price. However, even at present, most of our member countries construct their national IO tables at producer s price. ortunately, in recent years, much more Asian economies have been considering to establish or improve their SUT system under an international joint project supported by AD 5. If the SUT data for more Asian countries can be available, the national IO table at basic prices can be easily estimated. 3. OMPIATION O THE TRANSNATIONA INTERREGIONA INPUT-OUTPUT TAE ETWEEN HINA AND APAN Triggered by the commencement of Open-Door Policy in 9 and the entry into the WTO in 200, an increasing presence of hina has brought a significant change in the industrial network between hina and apan. Given the continuous reduction of international trade cost, the economic interdependence and integration between hina and apan experienced great development in recent years. At the same time, the cross-border economic interaction between the two countries by the way of production networks also spreads to each country s domestic economies at regional level. Since both hina and apan have relatively large economic scales, the economic size of their domestic province sometimes is larger than a middle-size country. This fact attracts us to considering the possibility to provide some information which can be used to measure the transnational interregional economic impacts between hina and apan. On this account, IDE-ETRO, with its 40 years of experience in compiling the international input-output tables, launched a new international joint project in 2006 on the compilation of the first-ever Transnational Interregional Input-Output (TIIO Table between hina and apan, namely linking the individual interregional IO tables of the two countries. This section provides some fundamental information about the TIIO international project 3. Overview of the TIIO table The 2000 TIIO table between hina and apan is designed to depict the transnational, interregional and inter-industrial network extended over the following nineteen regions, namely, seven domestic regions of hina, eight domestic regions of apan, five ASEAN5 countries (Indonesia, Malaysia, the Philippines, Singapore and Thailand, two economies of East Asia (Korea and Taiwan and the United States. The domestic regions of hina and apan are from the aggregations of their domestic provinces. Annex 3 shows the detailed information of regions. Due to the time and budget limitation, the TIIO project was designed to be completed within one-year. or the simplicity and because of the data availability, relatively rough industrial classification (ten sectors is used in this project. onsidering the recent evolution of production technologies and the features of production networks, instead of 5 See Hashiguchi (20 for detailed survey based information on AD s technical assistance for adopting the Supply-Use framework in Asian economies. 6

17 using the traditional classification, such as light industry, heavy industry, the following ten sectors are employed in the TIIO table. 00 Agriculture, livestock, forestry and fishery 002 Mining and quarrying 003 Household consumption products (ife-related manufacturing products 004 asic industrial materials (Primary makers s manufacturing products 005 Processing and assembling (Secondary makers s manufacturing products 006 Electricity, gas and water supply 00 onstruction 00 Trade 009 Transportation 00 Services The layout of the 2000 TIIO table is shown in Annex 4, which is similar to the AIIO table as shown in Annex. It is easy to see that this table provides detailed transnational, interregional and inter-industrial information on the input composition and output distribution of each domestic industry vis-à-vis home as well as foreign countries industries. 3.2 Organizations involved in the TIIO project The original idea of compiling the TIIO table is from Nobuhiro Okamoto, a previous research fellow of IDE. ased on his idea, in 2006, IDE launched the TIIO international project. This project is mainly based on the collaboration of the following institutions. hina: State Information enter (SI apan: Applied Research Institute, Inc. (ARI apan: IDE-ETRO (project manager In addition, some experts from National Institute for and and Infrastructure Management (NIIM, Ministry of and, Infrastructure and Transport of apan, Pacific onsultants o,. TD. also involved in this project as consultants and advisors. 3.3 ompilation procedure and methodology of the TIIO table Roughly speaking, the compilation process of the TIIO table is very similar to the AIIO tables. There are mainly three steps. The first step is to adjust the format of different data source provided by hina and apan sides. The second step is to prepare the transnational and interregional trade data and split the import matrix of hina and apan s domestic interregional IO table by origins of the nineteen regions. inally, the third step is to link hina and apan s parts completed in the second step and make the balancing and reconciliation work.

18 3.3. Data collection and estimation The main data source used in the TIIO project is shown below: ( The 99 interregional IO table of hina (2 The 2000 interregional IO table of apan (3 The 2000 Asian international IO table (4 The 2000 commodity flow data by port between hina and apan (5 Other supplementary data The 2000 AIIO table is used as the control total data in the project. The SI and ARI are in charge of the compilation of hina s part and apan s part respectively. oth parts are the interregional IO tables with separate import matrix by nineteen origins (endogenous regions. In order to making these parts, both institutions use the information of commodity flow by port between hina and apan to split the import matrix of their interregional IO tables. The commodity flow data by port is estimated by our consultants coming from the NIIM ompilation methodology The 2000 TIIO table between hina and apan is compiled from the 2000 AIIO table and interregional IO tables of both hina and apan. Since there are not enough survey data for us to estimate the detailed transnational and interregional commodity flow, some relative strict assumptions are used. The main compilation method is based on the following principle. Suppose that there is an international (bilateral IO table between country R and country S and an interregional IO table between region p and region q for country R, the part of country R in the international IO table is divided into two regions (p and q by using information obtained from the interregional IO table and transnational commodity flow data by port. In order to divide the row-wise transactions of country R into regions, the regional ratio of interregional IO table is applied to the domestic transaction and the regional export ratio is used for dividing the export matrix to country S. Thus, the output structure of region p in country R can be estimated as follows: pp ij z pp ij pq ij ( z z z z pp il pp il pq il qp ij qp il qq ij f ( f f f f e ( e e ps i ps i e ( e e ps i ps i qs i qs i z RS ij f : RS il : z qq il RR ij f : intermediate output of region p in country R RR il : final demand of region p in country R exports from region p in country R to intermediate sectors in country S exports from region p in country R to final demand sectors in country S

19 p xi RR x p q i : gross output of region p in country R ( xi xi where, e represents exports. Region q in country R is also estimated in the same manner. After splitting the row vector of country R into regions, the column-wise division is done in the same way as described above. The input structure of region p in country R is estimated as follows: Sp mi ( m m Sp i m ( m m Wp i Wp i p vkj p q kj vkj ( v p j q j Sq i Wq i v z z R kj SR ij WR ij : imports of region p in country R from country S : imports of region p in country R from the rest of the world (W : value added of region p in country R p x j R x j : gross input of region p in country R ( x x where, m indicates imports. inally, region p s import matrix from country S is divided by region of origin in country S. et the region of country S be y, then we have e yr i yr ei y Sp mi ( m m Sp i Sq i z SR ij : imports of region p in country R from region y in country S The imbalances arise in both rows and columns. The adjustment based on the RAS method and the specialists opinions are performed to eliminate the discrepancies. 3.4 Some new movements towards the next joint TIIO projects After the Great East apan Earthquake happened on March 20, transnational economic impacts focusing on specific domestic regions have attracted a lot of attention for regional governments. Given the deepening economic interdependence and the increasing transnational spillover impact at domestic regional level among East-Asian economies, IDE is planning to launch a new international project of 2005 TIIO table involving the three East Asian countries, apan, hina and Korea from apan s fiscal year of 202. This will be also the first-ever joint project that links the domestic interregional/multiregional IO table together for the three East-Asian countries. The good news is that the 2005 interregional IO tables for apan and Korea have been available for us; the SI of hina is on the process of compiling the 2002 and 200 interregional IO tables which are expected to be officially published soon. This section provides some new information about the on-going compilation process of hina s multi-regional IO table and some related employment data. 9

20 3.4. ompilation of hina s multi-regional IO tables SI and IDE jointly compiled 99 hina -region, 30-industry multi-regional IO (MRIO model (see IDE 2003, which was widely used internationally. However, since 99, as hina s interregional trade and economic linkage strengthens, it s more urgent to update and develop 2002 and 200 MRIO model according to the compiling year of IO table in hina. ased on summarizing the experience on our 99 MRIO model development, SI further study and improve the development methods and procedures on MRIO model, then 2002 and 200 MRIO model are successively developed based on more normative method. In construction of 2002 and 200 MRIO models, we adopt the industrial classification used by the National ureau of Statistics (NS. We aggregate some of the service industries, so that the models are classified at 29-industry level. This industrial classification methodology is basically comparable to that used for the MRIO model for 99, allowing for the establishment of a series of MRIO models. The regional division for the 2002 and 200 models is identical to that for the 99 model. All research and development work for the 2002 and 200 MRIO models, however, is based on separate information on 30 provinces (municipalities, autonomous regions 6, allowing for different combinations in regional division and thus the establishment of MRIO models for different regions. There are several characteristics in our 2002 and 200 MRIO model development: first, the entire development is based on 30 provinces, various regional versions of MRIO model can be made after aggregating different province groups, according to needs of regional plan and policy design; second, in MRIO core work - estimating interregional trade coefficient, we propose specific models and estimation methods, based on entropy maximizing and gravity models; third, in cooperation with NS, we modify interregional transaction matrix, adopting the basic survey data reflecting inter-province inflow and outflow in National Input-Output Survey 200 ; fourth, when conducting balance adjustment, each province s table are fully used in total control that the sum of all provinces parts equals national table. During the compilation of the 2002 and 200 MRIO tables, the following difficulties and challenges have been our main targets for adjustment. (a Inconsistence between the trade data in national IO table and customs trade data Although the IO tables for provinces are improving continuously in term of the details of trade data, they still need to be further divided and be converted to import-inflow non-competitive input-output tables. During the adjustment of two-column trade data 6 Due to IO table of Tibet province is not compiled, Tibet is not included in provinces described in this paper. Also, economic structure of Tibet is not considered in the development of MRIO model. 20